1. Apple’s Platform Dominance Reassessed in the Agentic AI Era (BofA)

• Core Source

“In the Agentic AI era, value shifts away from ‘the best model’ toward the endpoint that controls user intent, personal context, app access, payments, authentication, permissions, and trust.”

“The price target has been raised from the existing $330 to $380. The new target is based on C27E EPS of $10.29 with a multiple of 37x.”

“The net incremental revenue opportunity on an F30E basis is estimated at $15–30 billion in the base scenario and $40–65 billion in the bull scenario, with potential EPS contribution of up to $2.”

“Services gross margin is 75.4%, far higher than Products at 36.8%, and Services accounts for only 26.2% of total revenue but 42.2% of total gross profit.”

“Complex Agentic workflows can reach 3,000–12,000 tokens, with BofA’s weighted average request volume at approximately 1,900 tokens.”

• Expected Impact

BofA’s report directly challenges the market’s prevailing perception of Apple as an AI laggard. The core thesis is that in the Agentic AI era, value does not accrue to the company with the best model, but to the endpoint platform that integrates user intent, payments, authentication, and permissions into a single trust ecosystem. The smartphone is that endpoint, and Apple is the only company that bundles iPhone, Apple Silicon, iOS, Apple Pay, Face ID, and the App Store into one cohesive architecture.

The monetization pathways are also laid out in concrete terms. BofA identifies six revenue streams: paid Apple Intelligence Pro subscriptions, model routing fees, App Store agent commerce take rate, Apple Pay and Wallet routing fees, Siri in-result advertising, and iCloud+ AI tiers. After netting out overlaps, the net incremental revenue opportunity by F30E is estimated at $15–30 billion in the base scenario and $40–65 billion in the bull scenario, with potential EPS contribution of up to $2.

Apple’s current revenue structure underscores why this matters. As of FY25, Services revenue accounts for only 26% of total revenue, yet bears 42% of total gross profit, with a gross margin of 75.4% versus Products at 36.8%. Should Agentic AI intensify the monetization of Apple’s installed base, the profit leverage effect would far exceed the revenue growth itself.

The critical execution risk lies in whether Apple can evolve Siri into a genuine orchestration layer. The first test will come at WWDC in June, where the market will look for concrete signs of progress on Agentic Siri capabilities.

2. NASA Unveils Three-Phase Lunar Base Roadmap — A Tale of Winners and Losers in the Moon Race

• Core Source

“NASA plans to invest approximately $30 billion over 11 years, with 79 missions planned. Based on this, lunar base construction and crew rotation will begin in 2030, with permanent crewed habitation capability to be secured from 2033.”

“Blue Origin and Firefly Aerospace each secured new contracts worth $1.9 billion and $75 million respectively.”

“For Intuitive Machines, which had been confident of winning the LTV contract, this was a painful defeat, and the company’s stock fell sharply as the news broke.”

“Of the CLPS 1.0 contracts announced to date, there are 11 in total, and given NASA’s stated Phase 1 target of 21 total landings, there is a high likelihood that approximately 10 additional CLPS 1.0 contracts will be announced.”

“NASA is investing more than $439 million in lunar terrain vehicle development, and the existing Lunar Gateway space station plan has reportedly been abandoned.”

• Expected Impact

NASA’s disclosure of its lunar base construction plan is not merely a space exploration headline — it is the first time the concrete contours of a civilian commercial space procurement market have come into view. Three points stand out.

First, the scale of the market has become visible. Eleven years, $300 billion, and 79 missions confirm to private space companies that a multi-year procurement pipeline is real. Notably, NASA officially abandoned the Lunar Gateway plan and pivoted to a more pragmatic roadmap centered on a lunar south pole outpost, which raises the credibility of actual budget execution.

Second, this round of contract awards reshuffled the competitive landscape. Intuitive Machines was eliminated from the LTV contract, with Astrolab and Lunar Outpost selected instead — a result that caught the market off guard and was immediately reflected in a sharp drop in LUNR’s stock. Firefly Aerospace, meanwhile, secured a $75 million MoonFall drone transport contract from NASA JPL, with Q1 revenue of $80.9 million (up 40% quarter-on-quarter) and full-year guidance of $420–450 million. Its stock surged more than 18% intraday on the news.

Third, the competition is far from over. With only 11 of the Phase 1 target of 21 landings covered by existing CLPS 1.0 contracts, approximately 10 more contract announcements are expected. CLPS 2.0 RFPs are also underway. Both LUNR and Firefly remain the only private companies to have successfully landed spacecraft on the lunar surface, meaning the head-to-head competition for lunar lander contracts between the two continues. Investors should view this round’s results not as a final verdict, but as a mid-game score in a multi-round contest.

3. Starlink Forces 5x Price Hike on U.S. Department of Defense — Weaponized Asset Status Elevates Musk’s Negotiating Power

• Core Source

“SpaceX executives argued that the military was paying approximately $5,000 per terminal in access fees while actually using approximately $25,000-level premium services.”

“SpaceX raised the cost of Starlink connectivity for drones by 5x to $25,000, and the U.S. Department of Defense ultimately agreed to pay the increased rate.”

“Starlink has become an essential asset in U.S. military operations, and Musk’s negotiating leverage and influence over the Department of Defense continues to grow.”

“B2B/B2G revenue accounted for approximately 36.7% of Starlink’s total revenue in ’25, rising steadily from 27.2% in ’23 and 36.4% in ’24.”

• Expected Impact

This episode is a defining case study demonstrating that Starlink’s business model is transitioning from a B2C satellite internet service to a mission-critical military and government infrastructure asset.

From a pricing power perspective, B2B/B2G customers face high switching costs and are therefore far less resistant to price increases than retail consumers. The moment U.S. military operations — including guidance of LUCAS loitering munitions and circumvention of Iranian communications blackouts — became dependent on Starlink in active combat, SpaceX was able to convert its supplier advantage into an actual price increase. The fact that a 5x hike from $5,000 to $25,000 per terminal was accepted outright is the proof.

Structurally, this is also a meaningful trend. Starlink’s B2B/B2G revenue share has risen from 27.2% in 2023 to 36.4% in 2024 and 36.7% in 2025, a steady structural expansion. As this share grows, the average unit price and profitability of Starlink’s overall revenue improves, since government and enterprise customers prioritize performance over cost in a way that retail subscribers do not.

For SpaceX, which is preparing for what could be one of the largest IPOs in history, the stability and high-margin profile of military and government contracts serves as a premium valuation factor. The key risk, however, is that escalating tensions with the DoD could trigger pressure to renegotiate contracts or accelerate the development of alternative satellite network capabilities.

4. NVIDIA Effectively Abandons Rubin CPX Roadmap — Strategic Pivot to Groq LPU-Centric AI Inference Architecture

• Core Source

“NVIDIA has effectively cancelled memory and substrate orders related to Rubin CPX.”

“In particular, there are no visible order movements for GDDR7, which had been cited as the target memory, and related development requests are also absent.”

“NVIDIA removed Rubin CPX from its roadmap at GTC in March, and brought Groq’s LPU — with which it signed a license agreement late last year — to the forefront.”

“The role of CPX has been absorbed into the Vera Rubin + Groq LPX + CMX combination.”

“Samsung Electronics is currently producing the Groq LP30 chip, with shipments scheduled for the second half of 2026.”

• Expected Impact

NVIDIA’s strategic pivot is being read as a signal of a shift in the dynamics of the AI inference chip market.

NVIDIA removed Rubin CPX from its roadmap at GTC in March. CPX had been conceived as a chip dedicated to the prefill bottleneck in the LLM inference process, but no official reason for its cancellation has been disclosed. NVIDIA brought Groq’s LPU — with which it signed a license agreement late last year — to the forefront, and the industry now views the role of CPX as having been absorbed into the Vera Rubin GPU + Groq LPX + CMX combination. Rather than Groq LPX fully replacing HBM, the structure is one where the SRAM-based LPU supplements the low-latency decode workloads where HBM GPUs are weaker.

The supply chain impact is immediate. The CPX-related orders that the GDDR7 memory and substrate industries had been counting on have evaporated. The memory industry has lost its GDDR7 supply expansion opportunity, while the substrate industry has lost the anticipated demand for GDDR7 memory substrates. Samsung Electronics, on the other hand, is directly producing the Groq LP30 chip with shipments scheduled for the second half of 2026, opening a new beneficiary pathway from a foundry perspective. The impact on HBM demand itself is viewed by the industry as limited, for the structural reasons outlined above.

5. EU Pursues Record DMA Fine Against Google Search Self-Preferencing — Big Tech Platform Regulation Enters a New Phase

• Core Source

“The European Commission is preparing to impose a large fine on Google for alleged violations of the Digital Markets Act (DMA), as reported by Handelsblatt.”

“The fine is reportedly being discussed at a level of ‘high three-digit millions of euros’, i.e. in the high hundreds of millions of euros range.”

“If confirmed, this would be the largest sanction ever imposed by the EU on the basis of DMA violations.”

“Handelsblatt notes that this could be the first time the EU has imposed a fine on Google on the basis of the DMA.”

“Handelsblatt notes that this measure could again strain transatlantic relations following the recent U.S.-EU tariff agreement.”

• Expected Impact

The significance of this case lies in the fact that the DMA has begun to function as a regulation with real enforcement teeth. The DMA, introduced by the EU in 2023, designates large platform operators — including Google, Meta, Apple, Amazon, and Microsoft — as “gatekeepers” and prohibits them from self-preferencing their own services. This fine would be both the first ever imposed on Google under the DMA and the largest in DMA history.

For investors, there are two core risks. First, structural pressure on Google’s search business model. If the EU’s demand to eliminate self-preferencing in search results is enforced, the traffic and revenue of Google’s vertically integrated search products — Google Shopping, Google Maps, Google Travel — face direct headwinds. Google has already applied changes to its search UI in Europe, but the EU has deemed these insufficient. Second, the transatlantic relationship variable. With the U.S.-EU relationship having partially recovered following a recent tariff agreement, this enforcement action confirms that Europe’s regulatory posture toward American big tech remains firmly in place.

Over the longer term, if the EU’s DMA enforcement intensifies, this case sets a precedent for similar actions against other designated gatekeepers — Meta, Apple, and Amazon among them. The official announcement is expected before the EU’s summer recess, and the confirmed fine amount and any additional remedies will be the key market focal points.

6. Qualcomm Secures AI Server ASIC Supply Contract with ByteDance — First Major Foothold Beyond Smartphones into Data Centers

• Core Source

“Qualcomm surges 5% on news of a contract to supply hundreds of millions of AI server processors to ByteDance. This appears to be an important first step in expanding from a mobile processor-centric portfolio to servers, much like NVIDIA.”

“ByteDance plans to purchase millions of AI-dedicated ASIC chips, to be used for AI agent software and data center infrastructure operations.”

“This contract is expected to be one of the first major customer wins for Qualcomm’s AI ASIC business. Qualcomm has historically had high dependence on its smartphone chip business and limited presence in the AI accelerator market.”

• Expected Impact

Qualcomm is the dominant player in the global smartphone AP market, but has been virtually absent from the AI data center semiconductor space. This ByteDance contract represents the first publicly disclosed instance of Qualcomm securing a major customer for AI server ASICs.

In terms of market structure, the AI chip market is currently dominated by NVIDIA GPUs, but large cloud and internet companies are rapidly expanding their adoption of custom or third-party ASICs to reduce costs and diversify supply. Google TPU, Amazon Trainium and Inferentia, and Meta’s MTIA are all part of this trend. ByteDance similarly has clear demand for custom ASICs to handle the massive AI recommendation algorithm and content analysis workloads of TikTok and Douyin.

For Qualcomm, this contract carries significance beyond the revenue itself — it establishes a reference case as an AI data center ASIC supplier. Qualcomm’s CEO has already publicly outlined a strategy to expand into CPU, AI inference accelerator, and custom ASIC markets, and this contract represents the first tangible result of that strategy. That said, the exact quantity implied by “millions of units” remains unspecified, and Qualcomm’s ability to demonstrate technical differentiation against established AI chip players such as NVIDIA, Broadcom, and Marvell is a challenge that now begins in earnest. For now, this is best interpreted as clearing a first gate, not winning the race.

7. Explosive Growth in the AI Inference Infrastructure Startup Market — Baseten ARR Up 20x Year-on-Year, Together AI Surpasses $1 Billion

• Core Source

“Baseten is currently in discussions to raise up to $1 billion at a valuation of $11 billion — more than double the $5 billion valuation from three months ago.”

“Baseten’s ARR stands at $600 million as of end of Q1, up approximately 20x year-on-year and 3x from $200 million at the start of the year.”

“Modal recently closed funding at a valuation of $4.65 billion, with ARR growing 5x since last autumn to approximately $300 million.”

“Together AI’s valuation is $8.5 billion, with ARR surpassing $1 billion as of end of Q1.”

“Fireworks AI is in discussions for new funding at a valuation of $15 billion — up nearly 4x from its $4 billion valuation in October. ARR is approximately $280 million.”

• Expected Impact

The proliferation of open-source AI models and the rapid expansion of agent applications are giving rise to a new platform layer market dedicated to AI inference infrastructure. Players such as Baseten, Modal, Together AI, and Fireworks AI have evolved well beyond simply renting GPU servers — they now provide integrated offerings spanning model serving, fine-tuning, inference operations, and workflow optimization.

Their growth rates are exceptional. Baseten’s ARR grew approximately 20x in a single year to reach $600 million, while Together AI has crossed the $1 billion ARR threshold. Valuations have followed suit, with Fireworks AI nearly quadrupling from $4 billion to $15 billion in six months, reflecting the market’s re-rating of AI inference infrastructure from commodity to core platform.

From an investment perspective, the growth of this market also serves as an indirect indicator of sustained NVIDIA GPU demand. As inference infrastructure startups scale, GPU demand broadens across a wider customer base. However, the structural risks these companies face are equally clear: rising GPU rental costs, the threat of major cloud providers (AWS, GCP, Azure) offering comparable services directly, and the paradox that as inference efficiency improves, per-unit revenue may compress. The most important shift is that the market has begun to view these companies as core platform-layer players in the AI inference era — and whether that perception holds will be the central variable determining their valuations going forward.